PerplexityBot fetched 60 of our 179 URLs in two bursts of 9 and 20 minutes, and 4 of our 5 newest posts inside 21 hours. Googlebot fetched 14 paths in the same 29 hours and none of the 5

Method, stated before the numbers

Every request to askedabout.com from a bot-shaped user agent is written to a ledger with the path, the vendor and kind the shipped classifier assigns, and a timestamp; our own probe rows are tagged and excluded from every count here. This read pulled the seven days to 09-07 16:49 UTC: 3279 rows, 78 of them probes, 3201 net, under the pull's 5000-row cap so nothing was dropped. A burst is a run of PerplexityBot rows with no gap over 60 minutes. A sweep is any burst of 10 or more rows; the sweep under study is every such burst on 09-06 and 09-07, and the comparison window runs from its first row to the read, 28.5 hours. Per-post counts come from the uncapped per-path index since each post's ship commit, with vendors joined from the capped pull; the two counts are asserted equal for every post. The citation gauge is the weekly 12-question, 4-engine, 3-sample self-audit, read from its stored answers.

The computation is the committed script `content/news/perplexity_sweep_cut.mjs`; it writes `content/news/perplexity-sweep-cut-2026-09-07.json` and this post's text is generated from that file. 21 of 21 self-checks pass, including the cap, the newest-row age, the per-post count agreement and the sitemap membership of every swept path.

PerplexityBot's seven days: 118 rows, 78 distinct paths, three sweeps

Two shapes. Eight of the 11 bursts are two rows: robots.txt plus one page, most days at least once. The other three are the sweeps: 37 rows across two bursts on 09-01 (36 paths), 44 rows in 9 minutes on 09-06, 17 rows in 20 minutes on 09-07. The 09-06 and 09-07 bursts overlap on one path only, robots.txt; together they are 60 distinct paths, 59 of them in the sitemap (33% of 179): /news 27, /robots.txt 1, / 1, /library 1, /about 1, /for 5, /guide 9, /data 7, /compare 6, /agency-5-pack 1, /terms 1. Nothing on our side changed between the two sweeps that would ask for one: PerplexityBot fetched the sitemap 0 times in the seven days, so the sweeps do not start from it.

Every crawler in the same 28.5 hours

Rows and reach are different numbers. Meta's crawler logged 573 rows on 72 paths, re-fetching the same pages; PerplexityBot logged 63 rows on 60 paths, one request per page. Amazonbot took the same shape on a smaller scale (34 rows, 34 paths). Googlebot's 39 rows are 14 paths: robots.txt 18 times, the sitemap 3 times, 9 /news posts (all published before 2026-09-05), one /guide, one /compare and one /data page.

The 5 newest posts: PerplexityBot 4, Googlebot 0

The four posts live for more than a day were each fetched once by PerplexityBot, inside the sweep, 14.4 to 20.4 hours after publish; the sweep did the new-page discovery, not a separate fetch. The newest post had been live 2.8 hours at the read and had 2 rows, YandexBot and ClaudeBot. Googlebot has 0 rows on the 5. This repeats the 09-01 read, where four AI crawlers fetched a new post inside nine hours and Googlebot had not fetched it in seven days.

Googlebot's one new index this run, read the same way

On the 14-URL index we pre-registered on 08-28, day 11: 3 of 14 indexed, verdict "INCONCLUSIVE — 3 of 14 indexed, under the bar of 7 and above 0". The new one is `what-queries-does-google-show-a-small-ai-search-site-for-83-of-2…`: Search Console lastCrawlTime 2026-09-07T12:42:09Z, matching a Googlebot row in our ledger to the second, and "Submitted and indexed" at the 16:39 UTC read, under four hours later. That row is the path's first Googlebot row since the index shipped on 08-28, and it is the second post on this index to go from its first Googlebot fetch to indexed inside a day, after yesterday's. Over the same seven days PerplexityBot fetched 5 of the 14 and Googlebot 3; the 11 not indexed have 0 Googlebot rows in seven days.

The citation gauge the same day, beside the sweep and not scored

The weekly gauge ran 5.4 hours after the sweep's last row. Perplexity cited us in 3 of 36 answers, all on `/compare/best-profound-alternatives`, the same count and the same page as the 08-31 run. That page was in the sweep (3 PerplexityBot rows in seven days). 60 paths fetched, 1 path cited, and the cited set did not move. We report the two side by side because they land on the same day; a 12-question panel cannot attribute a citation to a crawl, and this post does not.

What this means for a business that wants to be found by AI engines

• Perplexity refreshes its copy of a small site in whole-site bursts days apart, not page by page. A new page here was in PerplexityBot's log within 21 hours because a sweep happened to land; between sweeps the crawler fetched two paths a day. If your page is new, the question is when the next sweep runs, and our ledger says 5.3 days between the last two. How Perplexity picks what it cites is a separate question from whether it has your page.
• Being fetched is not being cited. 60 of our paths were fetched and the same one page was cited 3 of 36 times the same afternoon, as it was the week before. Crawler rows are evidence of access, not of visibility; the only way to know what an engine says about your business is to ask it, on a schedule. You can run that check on your own business.
• Googlebot is the outlier on a small site. 39 rows in 28.5 hours, 18 of them robots.txt, none on the 5 newest posts, while at least 5 vendors fetched each of the four day-old posts inside 24 hours. When Google does fetch a post here, it has indexed it the same day twice running; the wait is for the fetch.

Limits

• Vendor is the user-agent string as classified; rows are not reverse-DNS verified. PerplexityBot's declared token was `PerplexityBot/1.0` on every row.
• The 60-minute burst rule and the 10-row sweep floor are this post's definitions, chosen before the tables were drawn; the 09-07 burst counted as part of one sweep with 09-06 because they overlap on robots.txt only and together cover the site.
• One site, one week, three sweeps. 5.3 days between sweeps is one interval, not a cadence.
• The window (28.5 hours) starts at the sweep's first row, so it is drawn around PerplexityBot; every other vendor's count is over the same clock, not over a window drawn for it.
• The user-agent pull is capped at 5000 rows and returned 3279; the per-post totals come from the uncapped index and agree.
• The gauge comparison is two runs, seven days apart, on one page. Nothing here scores the sweep's effect on citations; the pre-committed read below does.

Trailing indexation, read the same run

The 8 posts published 2026-09-04 to 09-07 were inspected in the same run: 5 "Discovered, not indexed", 3 unknown to Google, 0 indexed, 0 with a lastCrawlTime, 0 Googlebot rows on their paths since publish against 87 rows from other crawlers. PerplexityBot has fetched 7 of the 8.

The next reading, pre-committed

2026-09-14: The 2026-09-14 13:07 UTC gauge run (run 7): Perplexity's answers citing askedabout.com out of 36, and the cited paths, reported beside whether PerplexityBot swept again in the seven days between (a sweep = a burst of 10 or more rows). Pre-registered falsification: if a second sweep lands and Perplexity's cited-path set is unchanged from run 6, sweeps are not adding cited pages on this panel; if the set grows without a sweep, the sweep is not the mechanism either. Also 2026-09-11: the pre-registered 14-URL hub read (bar 7 indexed). Published either way.